Eco and Workarounds For Bugs in Esp32 en

ESP32
ECO and Workarounds for Bugs
Version 2.3
Espressif Systems
Copyright © 2020
www.espressif.com
About This Guide
This document details hardware errata and workarounds in the ESP32.
Release Notes
Date Version Release notes
2016-11 V1.0 Initial release.
2016-12 V1.1 Modified the MEMW command in Section 3.2.
Changed the description of Section 3.1;

2017-04 V1.2
Added a bug in Section 3.8.
2017-06 V1.3 Added items 3.9 and 3.10
2018-02 V1.4 Corrected typos in the register names in Section 3.3.
2018-02 V1.5 Added Section 3.11.
2018-05 V1.6 Overall update.
2018-05 V1.7 Added Section 3.12.
2018-12 V1.8 Added Section 3.13: ESP32 CAN Errata.
• Added chip revision 3 in Table 1-1

• Added note of fixes in sections 3.9 and 3.10
2020-03-16 V1.9 • Added Section 3.13.10
• Added Section 3.14
• Added documentation feedback link

2020-05-08 V2.0 • Added a note in Section 3.3
• Updated the address ranges of space A and B in Section 3.10 and
fixed a typo
2020-05-14 V2.1 Added a note of fix in Section 3.8.
2020-06-08 V2.2 Added sections 3.17 and 3.18.
Updated section 3.16, and provided more information about UART

2020-09-25 V2.3
FIFO read operation.
Documentation Change Notification
Espressif provides email notifications to keep customers updated on changes to

technical documentation. Please subscribe at www.espressif.com/en/subscribe. Note
that you need to update your subscription to receive notifications of new products
you are not currently subscribed to.
Certification
Download certificates for Espressif products from www.espressif.com/en/certificates. 
Table of Contents
1. Chip Revision .........................................................................................................................1
2. Errata List ...............................................................................................................................2
3. Errata Descriptions and Workarounds..................................................................................4

3.1. A spurious watchdog reset occurs when ESP32 is powered up or wakes up from Deep-sleep.
4
3.2. When the CPU accesses external SRAM through cache, under certain conditions read and
write errors occur. ......................................................................................................................4
3.3. When the CPU accesses peripherals and writes a single address repeatedly, some writes
may be lost. ................................................................................................................................5
3.4. The Brown-out Reset (BOR) function does not work. The system fails to boot up after BOR. .6
3.5. The CPU crashes when the clock frequency switches directly from 240 MHz to 80/160 MHz. 6
3.6. GPIO pull-up and pull-down resistors for pads with both GPIO and RTC_GPIO functionality
can only be controlled via RTC_GPIO registers. ........................................................................6
3.7. Audio PLL frequency range is limited. .......................................................................................7
3.8. Due to the flash start-up time, a spurious watchdog reset occurs when ESP32 is powered up
or wakes up from Deep-sleep. ...................................................................................................7
3.9. When the CPU accesses the external SRAM in a certain sequence, read & write errors can
occur. .........................................................................................................................................8
3.10. When each CPU reads certain different address spaces simultaneously, a read error can
occur. .........................................................................................................................................9
3.11. When certain RTC peripherals are powered on, the inputs of GPIO36 and GPIO39 will be
pulled down for approximately 80 ns. ........................................................................................9
3.12. When the LEDC is in decremental fade mode, a duty overflow error can occur. ......................9
3.13. ESP32 CAN Errata ...................................................................................................................10
3.13.1. Receive Error Counter (REC) is allowed to change whilst in reset mode or bus-off
recovery. .....................................................................................................................10
3.13.2. Error status bit is not frozen during bus-off recovery. ................................................10
3.13.3. Message transmitted after bus-off recovery is erroneous. .........................................11
3.13.4. Reading the interrupt register can lead to a transmit interrupt being lost. .................11
3.13.5. Receiving an erroneous data frame can cause the data bytes of the next received
data frame to be invalid. .............................................................................................11
3.13.6. After losing arbitration, a dominant bit on the 3rd bit of intermission is not interpreted
as an SOF. ...................................................................................................................12
3.13.7. When the 8th bit of the error delimiter is dominant, the error passive state is not
entered. .......................................................................................................................12
3.13.8. Suspend transmission is included even after losing arbitration. ................................12
3.13.9. When a stuff error occurs during arbitration whilst being transmitter, any errors in the
subsequent error/overload frame will not increase the TEC. ......................................13
3.13.10.A negative phase error where |e| > SJW(N) will cause the remaining transmitted bits
to be left shifted. .........................................................................................................13
3.14. The ESP32 GPIO peripheral may not trigger interrupts correctly. ...........................................13
3.15. The ESP32 chip may have a live lock under certain conditions that will cause interrupt
watchdog issue. .......................................................................................................................14
3.16. There are limitations to the CPU access to 0x3FF0_0000 ~ 0x3FF1_EFFF and 0x3FF4_0000
~ 0x3FF7_FFFF address spaces. .............................................................................................15
3.17. UART fifo_cnt is inconsistent with FIFO pointer. ......................................................................15
3.18. CPU has limitations when accessing peripherals in chips. ......................................................16

1. Chip Revision
!
1. Chip Revision
The chip revision is identified by the registers and eFuse identification bits. Details can be
found in the table below.
Table 1-1. Chip Revision
Register address
Chip revision APB_CTRL_DATE[31] EFUSE_BLK0_RDATA5[20] EFUSE_BLK0_RDATA3[15]
V0, without ECO 0 0 0
ECO, V1 0 0 1
ECO, V3 1 1 1
Espressif ! /!17
1
Submit Documentation Feedback 2020-09-25
2. Errata List
!
2. Errata List
Table 2-1. Errata Summary
Section Title Affected revisions
A spurious watchdog reset occurs when ESP32 is powered up or

Section 3.1 V0
wakes up from Deep-sleep.
When the CPU accesses external SRAM through cache, under

Section 3.2 V0
certain conditions read and write errors occur.
When the CPU accesses peripherals and writes a single address

Section 3.3 V0
repeatedly, some writes may be lost.
The Brown-out Reset (BOR) function does not work. The system fails
Section 3.4 V0
to boot up after BOR.
The CPU crashes when the clock frequency switches directly from
Section 3.5 V0
240 MHz to 80/160 MHz.
GPIO pull-up and pull-down resistors for pads with both GPIO and
Section 3.6 RTC_GPIO functionality can only be controlled via RTC_GPIO V0/V1/V3
registers.
Section 3.7 Audio PLL frequency range is limited. V0
Due to the flash start-up time, a spurious watchdog reset occurs

Section 3.8 V0/V1
when ESP32 is powered up or wakes up from Deep-sleep.
When the CPU accesses external SRAM in a certain sequence, read

Section 3.9 V1
and write errors can occur.
When each CPU reads certain different address spaces

Section 3.10 V0/V1
simultaneously, a read error can occur.
When certain RTC peripherals are powered on, the inputs of GPIO36
Section 3.11 V0/V1/V3
and GPIO39 will be pulled down for approximately 80 ns.
When the LEDC is in decremental fade mode, a duty overflow error

can occur.
Section 3.13 ESP32 CAN Errata
Receive Error Counter (REC) is allowed to change whilst in reset

Section 3.13.1 V0/V1/V3
mode or bus-off recovery.
Section 3.13.2 Error status bit is not frozen during bus-off recovery. V0/V1/V3
Section 3.13.3 Message transmitted after bus-off recovery is erroneous. V0/V1/V3
Reading the interrupt register can lead to a transmit interrupt being

lost.
Receiving an erroneous data frame can cause the data bytes of the
next received data frame to be invalid.
Espressif ! /!17
2
2. Errata List
!
Section Title Affected revisions
After losing arbitration, a dominant bit on the 3rd bit of intermission is

not interpreted as an SOF.
When the 8th bit of the error delimiter is dominant, the error passive
state is not entered.
Section 3.13.8 Suspend transmission is included even after losing arbitration. V0/V1/V3
When a stuff error occurs during arbitration whilst being transmitter,

Section 3.13.9 any errors in the subsequent error/overload frame will not increase the V0/V1/V3
TEC.
A negative phase error where |e| > SJW(N) will cause the remaining
transmitted bits to be left shifted.
Section 3.14 The ESP32 GPIO peripheral may not trigger interrupts correctly. V0/V1/V3
The ESP32 chip may have a live lock under certain conditions that will
Section 3.15 V3
cause interrupt watchdog issue.
There are limitations to the CPU access to 0x3FF0_0000 ~

0x3FF1_EFFF and 0x3FF4_0000 ~ 0x3FF7_FFFF address spaces.
Section 3.17 UART fifo_cnt is inconsistent with FIFO pointer. V0/V1/V3
Section 3.18 CPU has limitations when accessing peripherals in chips. V0/V1/V3
Espressif ! /!17
3
3. Errata Descriptions and Workarounds
!
3. Errata Descriptions and

Workarounds
3.1. A spurious watchdog reset occurs when ESP32 is powered up
or wakes up from Deep-sleep.
Workarounds:
When waking from Deep-sleep, this bug is worked around automatically in ESP-IDF V1.0
and newer.
During initial power-up the spurious watchdog reset cannot be worked around, but ESP32
will boot normally after this reset.
Workaround Details:
To work around the watchdog reset when waking from Deep-sleep, the CPU can execute a
program from RTC fast memory. This program must clear the illegal access flag in the
cache MMU as follows:
1. Set the PRO_CACHE_MMU_IA_CLR bit in DPORT_PRO_CACHE_CTRL1_REG to 1.
2. Clear this bit.
Fixes:
This issue is fixed in silicon revision 1.
3.2. When the CPU accesses external SRAM through cache, under
certain conditions read and write errors occur.
Details:
Access to external SRAM through cache will cause read and write errors if these operations
are pipelined together by the CPU.
Workarounds:
There is no automatic workaround available in software.
Workaround Details:
If accessing external SRAM from a revision 0 ESP32, users must ensure that access is
always one-way—only a write or a read can be in progress at a single time in the CPU
pipeline.
The MEMW instruction can be used: insert __asm__("MEMW") after any read from external
PSRAM that may be followed by a write to PSRAM before the CPU pipeline is flushed.
Espressif ! /!17
4 2020-09-25
Submit Documentation Feedback
!
Fixes:
3.3. When the CPU accesses peripherals and writes a single

address repeatedly, some writes may be lost.
Details:
Some ESP32 peripherals are mapped to two internal memory buses (AHB & DPORT).
When written via DPORT, consecutive writes to the same address may be lost.
Workarounds:
This issue is automatically worked around in the drivers of ESP-IDF V1.0 and newer.
Workaround Details:
When writing the same register address (i.e., FIFO-like addresses) in sequential instructions,
use the equivalent AHB address not the DPORT address.
(For other kinds of register writes, using DPORT registers will give better write
performance.)
Registers DPORT Addresses AHB (Safe) Addresses
UART_FIFO_REG 0x3FF40000 0x60000000
UART1_FIFO_REG 0x3FF50000 0x60010000
UART2_FIFO_REG 0x3FF6E000 0x6002E000
I2S0_FIFO_RD_REG 0x3FF4F004 0x6000F004
I2S1_FIFO_RD_REG 0x3FF6D004 0x6002D004
GPIO_OUT_REG 0x3FF44004 0x60004004
GPIO_OUT_W1TC_REG 0x3FF4400c 0x6000400c
GPIO_OUT1_REG 0x3FF44010 0x60004010
GPIO_OUT1_W1TS_REG 0x3FF44014 0x60004014
GPIO_OUT1_W1TC_REG 0x3FF44018 0x60004018
GPIO_ENABLE_REG 0x3FF44020 0x60004020
GPIO_ENABLE_W1TS_REG 0x3FF44024 0x60004024
GPIO_ENABLE_W1TC_REG 0x3FF44028 0x60004028
GPIO_ENABLE1_REG 0x3FF4402c 0x6000402c
GPIO_ENABLE1_W1TS_REG 0x3FF44030 0x60004030
Espressif ! /!17
5 2020-09-25
!
Registers DPORT Addresses AHB (Safe) Addresses
GPIO_ENABLE1_W1TC_REG 0x3FF44034 0x60004034
Fixes:
⚠ Notice:
Software cannot use AHB addresses to read FIFO.
3.4. The Brown-out Reset (BOR) function does not work. The
system fails to boot up after BOR.
Workarounds:
There is no workaround for this issue.
Fixes:
3.5. The CPU crashes when the clock frequency switches directly
from 240 MHz to 80/160 MHz.
Workarounds:
This issue is automatically worked around in ESP-IDF V2.1 and newer.
Workaround Details:
When switching frequencies, use intermediate frequencies as follows:
(1) 2 MHz <-> 40 MHz <-> 80 MHz <-> 160 MHz
(2) 2 MHz <->40 MHz <->240 MHz
Fixes:
3.6. GPIO pull-up and pull-down resistors for pads with both GPIO
and RTC_GPIO functionality can only be controlled via
RTC_GPIO registers.
Details:
For these pads, the GPIO pull-up and pull-down configuration register fields are non-
functional.
Espressif ! /!17
6 2020-09-25
!
Workarounds:
This issue is automatically worked around when using GPIO drivers in ESP-IDF V2.1 or
newer.
Workaround Details:
Use RTC_GPIO registers for both GPIO and RTC_GPIO functions.
3.7. Audio PLL frequency range is limited.

Details:
When configuring the Audio PLL, configuration registers sdm0 & sdm1 are not used. This
limits the range and precision of PLL frequencies which can be configured.
For chip revision 0, the Audio PLL frequency is calculated in hardware as follows:
fxtal(sdm2+4)
fout =
! 2(odiv+2)
For chip revision 1 onwards this bug is fixed and the Audio PLL frequency is calculated in
hardware as follows:
sdm1 sdm0
fxtal(sdm2+ 8 + + 4)
2 216
fout =
! 2(odiv+2)
Workarounds:
The particular hardware frequency calculation is automatically accounted for when setting
Audio PLL frequency via the I2S driver in ESP-IDF V3.0 and newer. However, the range and
precision of available Audio PLL frequencies is still limited when using silicon revision 0.
Fixes:
3.8. Due to the flash start-up time, a spurious watchdog reset

occurs when ESP32 is powered up or wakes up from Deep-
sleep.
Details:
If the ESP32 reads from the flash chip before it is ready, invalid data can cause booting to
fail until a Watchdog Timer reset occurs. This can occur on power-on and on wake from
Deep-sleep, if the ESP32 VDD_SDIO is used to power the flash chip.
Espressif ! /!17
7 2020-09-25
!
Workarounds:
1. Replace the flash chip with one with a fast start-up time (<800 μs from power-on to
ready to read). This works around the issue for both power-on and wake from Deep-
sleep.
2. When waking from Deep-sleep, this issue is automatically worked around in ESP-IDF
V2.0 and newer (the delay to wait can be configured if necessary). In this workaround,
the CPU executes from RTC fast memory immediately after waking and a delay is added
before continuing to read the program from flash.
Fixes:
This issue is fixed in silicon revision 3 (ECO V3).
3.9. When the CPU accesses the external SRAM in a certain

sequence, read & write errors can occur.
Details:
This error can occur when the CPU executes the following instructions to access external
SRAM:
store.x at0, as0, n
load.y at1, as1, m
In the pseudo-assembly instructions above, store.x represents an x-bit write operation,

while load.y represents a y-bit read operation. as0+n and as1+m represent the same
address in external SRAM.
• The instructions can be sequential or contained within the same pipeline (less than
four intermediate instructions, and no pipeline flushes.)
• When x>=y, the data write may be lost. (NOTE: when both the load and the store
refer to 32-bit values, the write is only lost if an interrupt occurs between the first and
second instructions.)
• When x <y, data writes may be lost and invalid data may be read.
Workarounds:
This bug is automatically worked around when external SRAM use is enabled in ESP-IDF
V3.0 and newer.
Workaround Details:
• When x>=y, insert four nop instructions between store.x and load.y.
• When x <y, insert a memw instruction between store.x and load.y.
Fixes:
Espressif ! /!17
8 2020-09-25
!
3.10. When each CPU reads certain different address spaces

simultaneously, a read error can occur.
Details:
Running in dual-core CPU mode, when one CPU bus reads address space A
(0x3FF0_0000 ~ 0x3FF1_EFFF), while the other CPU bus reads address space B
(0x3FF4_0000 ~ 0x3FF7_FFFF), an incorrect read can be generated on the CPU reading
address space B.
Workarounds:
This issue is automatically worked around in ESP-IDF V3.0 and newer.
Workaround Details:
Either of the following workarounds can be used:
• When either CPU reads address space A, prevent the other CPU bus from reading
address space B via locks and interrupts.
• Before reading address space A, disable interrupts and insert a read from address
space B on the same CPU (read a non-FIFO register, e.g., 0x3FF40078).
Fixes:
3.11. When certain RTC peripherals are powered on, the inputs of
GPIO36 and GPIO39 will be pulled down for approximately 80
ns.
Details:
Powering on the following RTC peripherals will trigger this issue:
• SARADC1
• SARADC2
• AMP
• HALL
Workarounds:
When enabling power for any of these peripherals, ignore input from GPIO36 and GPIO39.
3.12. When the LEDC is in decremental fade mode, a duty overflow

error can occur.
Details:
This issue may happen when the LEDC is in decremental fade mode and
LEDC_DUTY_SCALE_HSCHn is 1. If the duty is 2LEDC_HSTIMERx_DUTY_RES, then the next one
Espressif ! /!17
9 2020-09-25
!
should be 2LEDC_HSTIMERx_DUTY_RES – 1, however, the next duty is actually

2LEDC_HSTIMERx_DUTY_RES+1, which indicates a duty overflow error. (HSCHn refers to high-speed
channel with n being 0-7; HSTIMERx refers to high-speed timer with x being 0-3.)
For low-speed channels, the same issue may also happen.
Workarounds:
This issue is automatically worked around in the LEDC driver since the ESP-IDF commit ID
b2e264e and will be part of the ESP-IDF V3.1 release.
Workaround Details:
When using LEDC, avoid the concurrence of following three cases:
1. The LEDC is in decremental fade mode;
2. The scale register is set to 1;
3. The duty is 2LEDC_HSTIMERx_DUTY_RES or 2LEDC_LSTIMERx_DUTY_RES.
3.13. ESP32 CAN Errata

3.13.1. Receive Error Counter (REC) is allowed to change whilst in reset mode or bus-off
recovery.
Details:
When the CAN controller enters reset mode (e.g., by setting the RESET_MODE bit or due
to a bus-off condition) or when the CAN controller undergoes bus-off recovery, the REC is
still permitted to change. This can lead to the following cases:
• Whilst in reset mode or bus-off recovery, a changing REC can lead to the error status
bit changing which in turn could trigger the error warning limit interrupt.
• During bus-off recovery, an REC > 0 can prevent the bus-off recovery process from
completing.
Workarounds:
When entering reset mode, the CAN controller should set the the LISTEN_ONLY_MODE to
freeze the REC. The desired mode of operation should be restored before exiting reset
mode or when bus-off recovery completes.
3.13.2. Error status bit is not frozen during bus-off recovery.

Details:
When the CAN controller undergoes the bus-off recovery process, the controller must
monitor 128 occurrences of the bus free signal (11 consecutive recessive bits) before it can
become error active again. The number of bus-free signals remaining is indicated by the
transmit error counter (TEC). Because the error status bit is not frozen during bus-off
recovery, its value will change when the transmit error counter drops below the user-defined
transmit error warning limit (96 by default) thus trigger the error warning limit interrupt
before bus-off recovery has completed.
Espressif ! /! 17
10 2020-09-25
!
Workarounds:
When undergoing bus-off recovery, an error warning interrupt does not necessarily indicate
the completion of recovery. Users should check the STATUS_NODE_BUS_OFF bit to verify
whether bus-off recovery has completed.
3.13.3. Message transmitted after bus-off recovery is erroneous.

Details:
Upon completion of bus-off recovery, the next message that the CAN controller transmits
may be erroneous (i.e., does not adhere to CAN frame format).
Workarounds:
Upon detecting the completion of bus-off recovery (via the error warning interrupt), the CAN
controller should enter then exit reset mode so that the controller’s internal signals are
reset.
3.13.4. Reading the interrupt register can lead to a transmit interrupt being lost.
Details:
The CAN controller's interrupt signals are cleared by reading the INTERRUPT_REG.
However, if a transmit interrupt occurs whilst the INTERRUPT_REG is being read (i.e., in the
same APB clock cycle), the transmit interrupt is lost.
Workarounds:
When a message is awaiting completion of transmission (i.e., transmission has been
requested), users should also check the STATUS_TRANSMIT_BUFFER bit each time the
INTERRUPT_REG is read. A set STATUS_TRANSMIT_BUFFER bit whilst the
CAN_TRANSMIT_INT_ST is not indicates a lost transmit interrupt.
3.13.5. Receiving an erroneous data frame can cause the data bytes of the next received
data frame to be invalid.
Details:
When the CAN controller is receiving a data frame and a bit or stuff error occurs in the data
or CRC fields, some data bytes of the next received data frame may be shifted or lost.
Therefore, the next received data frame (including those filtered out by the acceptance filter)
should be considered invalid.
Workarounds:
Users can detect the errata triggering condition (i.e., bit or stuff error in the data or CRC
field) by setting the INTERRUPT_BUS_ERR_INT_ENA and checking the
ERROR_CODE_CAPTURE_REG when a bus error interrupt occurs. If the errata condition is
met, the following workarounds are possible:
• The CAN controller can transmit a dummy frame with 0 data bytes to reset the
controller’s internal signals. It is advisable to select an ID for the dummy frame that
can be filtered out by all nodes on the CAN bus.
Espressif ! /! 17
11 2020-09-25
!
• Hardware reset the CAN controller (will require saving and restoring the current
register values).
3.13.6. After losing arbitration, a dominant bit on the 3rd bit of intermission is not
interpreted as an SOF.
Details:
The CAN2.0B protocol stipulates that a dominant bit on the 3rd bit of intermission shall be
interpreted as a Start of Frame (SOF). Therefore, nodes shall begin receiving or transmitting
(i.e., competing for arbitration) the ID field on the next bit.
When the CAN controller loses arbitration and the following intermission’s 3rd bit is
dominant, the CAN controller will not interpret this as an SOF and will make no attempt to
compete for arbitration (i.e., does not retransmit its frame).
Workarounds:
3.13.7. When the 8th bit of the error delimiter is dominant, the error passive state is not
entered.
Details:
When the CAN controller is the transmitter and has a TEC value between 120 and 127,
transmitting an error frame will increment its TEC by 8 thus make the controller error
passive (due to TEC becoming >= 128). However, if the 8th bit of the error delimiter is
dominant, the TEC will still increment by 8 but the controller will not become error passive.
Instead, the controller will become error passive when another error frame is transmitted.
Note that the controller will still generate the required overload frame due to the dominant
8th bit.
Workarounds:
3.13.8. Suspend transmission is included even after losing arbitration.

Details:
The CAN2.0B protocol stipulates that an error passive node that was the transmitter of a
message shall add a suspend transmission field within the subsequent interframe space.
However, error passive receivers shall not add a suspend transmission field.
When the CAN controller is error passive and loses arbitration (hence becomes a receiver),
it will still add a suspend transmission field in the subsequent interframe space. This results
in the CAN controller being late to start retransmission. Therefore, if another node transmits
immediately after the interframe space is over, the CAN controller will fail to compete for
arbitration due to the other nodes not including a suspend transmission field in their
interframe space (as per CAN2.0B specification).
Espressif ! /! 17
12 2020-09-25
!
Workarounds:
3.13.9. When a stuff error occurs during arbitration whilst being transmitter, any errors in
the subsequent error/overload frame will not increase the TEC.
When a stuff error occurs during arbitration whilst being transmitter, the CAN2.0B protocol
stipulates that an error frame be transmitted but the TEC should not increase (Exception 2
of Rule 3). The CAN controller is able to fulfill these requirements without issue.
However, errors within the subsequent error/overload frames themselves will fail to increase
the CAN controller’s TEC. Therefore, when a stuff error occurs during arbitration whilst
being transmitter, the TEC will fail to increase in the following cases:
• Bit error in an active error flag or overload flag (Rule 4).
• Detecting too many dominant bits after the transmission of active error, passive error
flag, and overload flags (Rule 6).
Workarounds:
3.13.10.A negative phase error where |e| > SJW(N) will cause the remaining transmitted bits
to be left shifted.
Details:
When the CAN controller encounters a recessive to dominant edge with a negative phase
error (i.e., the edge is early), it will correct for the phase error using resynchronization as
required by the CAN2.0B protocol. However, if the CAN controller is acting as transmitter
and encounters a negative phase error where e < 0 and |e| > SJW, the bits transmitted
following the phase error will be left shifted by one bit. Thus, the transmitted frame's
contents (i.e., DLC, data bytes, CRC sequence) will be corrupted.
Workarounds:
3.14. The ESP32 GPIO peripheral may not trigger interrupts

correctly.
Details:
The ESP32 GPIO peripheral may not trigger interrupts correctly if multiple GPIO pads are
configured with edge-triggered interrupts.
Workaround 1:
Follow the steps below to trigger a GPIO interrupt on a rising edge:
1. Set the GPIO interrupt type to high.
2. Set the interrupt trigger type of the CPU to edge.
Espressif ! /! 17
13 2020-09-25
!
3. After the CPU services the interrupt, change the GPIO interrupt type to low. A second
interrupt occurs at this time, and the CPU needs to ignore the interrupt service
routine.
Similarly, follow the steps below to trigger a GPIO interrupt on a falling edge:
1. Set the GPIO interrupt type to low.
2. Set the interrupt trigger type of the CPU to edge.
3. After the CPU services the interrupt, change the GPIO interrupt type to high. A
second interrupt occurs at this time, and the CPU needs to ignore the interrupt
service routine.
Workaround 2:
Assuming GPIO0 ~ GPIO31 is Group1 and GPIO32 ~ GPIO39 is Group2.
• If an edge-triggered interrupt is configured in either group then no other GPIO
interrupt of any type should be configured in the same group.
• Any number of level-triggered interrupts can be configured in a single group, if no
edge-triggered interrupts are configured in that group.
3.15. The ESP32 chip may have a live lock under certain conditions
that will cause interrupt watchdog issue.
Details:
On ESP32 ECO V3, when the following conditions are met at the same time, a live lock will
occur, causing the CPUs to get stuck in the state of memory access and stop executing
instructions.
1. Dual-core system.
2. Of the four Instruction/Data buses (IBUS/DBUS) that access external memory, three
simultaneously initiate access requests to the same cache set, and all three requests
result in cache misses.
Workarounds:
When a live lock occurs, software proactively or passively recognizes and unlocks the
cache line contention, and then the two cores complete their respective cache operations
one after another, following a first-come, first-served policy, to resolve the live lock. The
detailed process is as follows:
1. If the live lock occurs when the instructions executed by the two cores are not in the
critical section of the code, the various types of system interruptions will proactively
release the cache line competition and resolve the live lock.
2. If the live lock occurs when the instructions executed by the two cores are located in
the critical section of the code, the system will mask interrupts at level 3 and below.
Therefore, software needs to set up a high priority (level 4 or 5) interrupt for each core
in advance, connect the interrupts to the same timer, and configure an appropriate
timeout threshold. The timer timeout interrupt generated by the live lock will force
Espressif ! /! 17
14 2020-09-25
!
both cores to enter the high-priority interrupt handler, thereby releasing the IBUS of
both cores to resolve the live lock. The live lock resolution process is completed in
three stages:
(a) In the first stage, both cores wait for the CPU write buffer to be cleared.
(b) In the second stage, one core (Core 0) waits and the other core (Core 1)
executes instructions.
(c) In the third stage, Core 1 waits and Core 0 executes instructions.
3.16. There are limitations to the CPU access to 0x3FF0_0000 ~

0x3FF1_EFFF and 0x3FF4_0000 ~ 0x3FF7_FFFF address
spaces.
Details:
1. The CPU read operations that fall in these two address spaces are speculative.
Speculative read operations can cause the behavior described by the program to be
inconsistent with the actual behavior of the hardware.
2. If the two CPUs continuously access address space 0x3FF0_0000 ~ 0x3FF1_EFFF
at the same time, some of the access may be lost.
3. When the CPU reads FIFO through the address space 0x3FF4_0000 ~
0x3FF7_0000, the FIFO read pointer is updated with delays. As the CPU frequency
increases, the interval between two consecutive FIFO reads initiated by the CPU is
shortened. When a new FIFO read request arrives, the FIFO read pointer has not
been updated, causing the CPU to read the value of the previous FIFO read
operation.
Workarounds:
1. Insert "MEMW" instruction before the CPU access operation that falls in these two
address spaces. That is, in C/C++, software needs to always use the "volatile"
attribute when accessing registers in these two address spaces.
2. When the CPU frequency is 160 MHz, add six “nop” between two consecutive FIFO
reads. When the CPU frequency is 240 MHz, add seven “nop” between two
consecutive FIFO reads.
3.17. UART fifo_cnt is inconsistent with FIFO pointer.

Details:
When software uses DPORT to read UART fifo_cnt and FIFO pointer, and such operation is
interrupted, FIFO pointer will not decrement, giving invalid fifo_cnt.
Workarounds:
When using DPORT to read FIFO, please calculate the number of bytes according to the
value of FIFO pointer.
Espressif ! /! 17
15 2020-09-25
!
Register Chip Version

Address space (Bus) Operation
type
V0 V1 V3
Write ✔ ✔
0x3FF0_0000 ~ Non-FIFO
0x3FF1_EFFF and Read ✖ (See 3.10) ✔
0x3FF4_0000 ~ Write ✖ (See 3.3) ✔

0x3FF7_FFFF (DPORT) FIFO
Read ✔ ✔
Write ✔
Non-FIFO
Read ✔
0x6000_0000 ~
0x6003_FFFF (AHB) Write ✔
FIFO
Read ✖ (No such feature, unpredictable results)
Legend
✔ - operation is executed correctly
✖ - operation fails
3.18. CPU has limitations when accessing peripherals in chips.

Details:
As described in Section 3.3, 3.10 and 3.16, CPU has limitations when accessing peripherals in chips
of different versions using 0x3FF0_0000 ~ 0x3FF1_EFFF, 0x3FF4_0000 ~ 0x3FF7_FFFF, and
0x6000_0000 ~ 0x6003_FFFF.
Espressif ! /! 17
16 2020-09-25
Disclaimer and Copyright Notice
Information in this document, including URL references, is subject to change without
notice.
THIS DOCUMENT IS PROVIDED AS IS WITH NO WARRANTIES WHATSOEVER,
INCLUDING ANY WARRANTY OF MERCHANTABILITY, NON-INFRINGEMENT, FITNESS
FOR ANY PARTICULAR PURPOSE, OR ANY WARRANTY OTHERWISE ARISING OUT
OF ANY PROPOSAL, SPECIFICATION OR SAMPLE.
All liability, including liability for infringement of any proprietary rights, relating to use of
information in this document is disclaimed. No licenses express or implied, by estoppel or
otherwise, to any intellectual property rights are granted herein.
The Wi-Fi Alliance Member logo is a trademark of the Wi-Fi Alliance. The Bluetooth logo is
a registered trademark of Bluetooth SIG.
All trade names, trademarks and registered trademarks mentioned in this document are
Espressif IoT Team
property of their respective owners, and are hereby acknowledged.
www.espressif.com Copyright © 2020 Espressif Systems (Shanghai) Co., Ltd. All rights reserved.

Eco and Workarounds For Bugs in Esp32 en

Uploaded by

Copyright:

Available Formats

Eco and Workarounds For Bugs in Esp32 en

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Eco and Workarounds For Bugs in Esp32 en

Uploaded by

Copyright:

Available Formats

ESP32

ECO and Workarounds for Bugs

Date Version Release notes

2016-11 V1.0 Initial release.

2016-12 V1.1 Modified the MEMW command in Section 3.2.

Changed the description of Section 3.1;

2017-06 V1.3 Added items 3.9 and 3.10

2018-02 V1.4 Corrected typos in the register names in Section 3.3.

2018-02 V1.5 Added Section 3.11.

2018-05 V1.6 Overall update.

2018-05 V1.7 Added Section 3.12.

2018-12 V1.8 Added Section 3.13: ESP32 CAN Errata.

• Added chip revision 3 in Table 1-1

• Added Section 3.15

2020-05-14 V2.1 Added a note of fix in Section 3.8.

2020-06-08 V2.2 Added sections 3.17 and 3.18.

Updated section 3.16, and provided more information about UART

Documentation Change Notification

Espressif provides email notifications to keep customers updated on changes to

2. Errata List ...............................................................................................................................2

3. Errata Descriptions and Workarounds..................................................................................4

3.7. Audio PLL frequency range is limited. .......................................................................................7

3.13. ESP32 CAN Errata ...................................................................................................................10

3.13.3. Message transmitted after bus-oﬀ recovery is erroneous. .........................................11

3.13.8. Suspend transmission is included even after losing arbitration. ................................12

3.17. UART fifo_cnt is inconsistent with FIFO pointer. ......................................................................15

3.18. CPU has limitations when accessing peripherals in chips. ......................................................16

Table 1-1. Chip Revision

Chip revision APB_CTRL_DATE[31] EFUSE_BLK0_RDATA5[20] EFUSE_BLK0_RDATA3[15]

V0, without ECO 0 0 0

Section Title Affected revisions

A spurious watchdog reset occurs when ESP32 is powered up or

When the CPU accesses external SRAM through cache, under

When the CPU accesses peripherals and writes a single address

Section 3.7 Audio PLL frequency range is limited. V0

Due to the flash start-up time, a spurious watchdog reset occurs

When the CPU accesses external SRAM in a certain sequence, read

When each CPU reads certain different address spaces

When the LEDC is in decremental fade mode, a duty overflow error

Section 3.13 ESP32 CAN Errata

Receive Error Counter (REC) is allowed to change whilst in reset

Section 3.13.3 Message transmitted after bus-off recovery is erroneous. V0/V1/V3

Reading the interrupt register can lead to a transmit interrupt being

Section Title Affected revisions

After losing arbitration, a dominant bit on the 3rd bit of intermission is

When a stuff error occurs during arbitration whilst being transmitter,

There are limitations to the CPU access to 0x3FF0_0000 ~

Section 3.17 UART fifo_cnt is inconsistent with FIFO pointer. V0/V1/V3

3. Errata Descriptions and

3.3. When the CPU accesses peripherals and writes a single

Registers DPORT Addresses AHB (Safe) Addresses

UART_FIFO_REG 0x3FF40000 0x60000000

UART1_FIFO_REG 0x3FF50000 0x60010000

UART2_FIFO_REG 0x3FF6E000 0x6002E000

I2S0_FIFO_RD_REG 0x3FF4F004 0x6000F004

I2S1_FIFO_RD_REG 0x3FF6D004 0x6002D004

GPIO_OUT_REG 0x3FF44004 0x60004004

GPIO_OUT_W1TC_REG 0x3FF4400c 0x6000400c

GPIO_OUT1_REG 0x3FF44010 0x60004010